Brain tissue presents some of the steepest challenges in single-cell genomics. Neurons are large, fragile, and entangled in dense processes that resist enzymatic dissociation. Whole-cell isolation from frozen or postmortem brain often yields biased cell populations — glia survive the protocol while neurons disappear — and the dissociation process itself can trigger transcriptional stress responses that muddy the biological signal. Single-nucleus RNA sequencing (snRNA-seq) sidesteps these problems by working with nuclei rather than intact cells, making it possible to profile the transcriptomes of neurons, glia, and other brain cell types from archived and postmortem specimens that would otherwise be inaccessible.

For neuroscientists studying neurodegeneration, brain tumors, or cell-type diversity across brain regions, snRNA-seq has become a core tool over the past decade. It works with flash-frozen surgical resections, banked postmortem tissue, and even samples stored for years at −80 °C — the very specimens that constitute the bulk of human brain collections worldwide. This article walks through what brain snRNA-seq can deliver — and where its limits lie — drawing on recent studies in Alzheimer's disease, Parkinson's disease, ALS, and brain tumor biology, along with practical guidance on region selection, cell-type annotation, and interpretation caveats.

Why Brain Researchers Turn to Nuclei

The practical case for snRNA-seq in brain research rests on three observations reinforced by systematic comparisons over the past five years.

1. Neuronal Recovery: A Decisive Advantage

Enzymatic dissociation — the standard approach for single-cell RNA-seq (scRNA-seq) — selectively damages or destroys large, morphologically complex neurons while enriching for smaller, hardier glial cells. When matched snRNA-seq and scRNA-seq are compared from the same cortical tissue sample, the difference in neuronal representation is striking:

For studies where neuronal transcriptomes are the primary interest, this bias alone makes snRNA-seq the more reliable choice.

2. Access to Archived Biospecimens

Many of the most valuable brain specimens exist only as frozen or postmortem tissue:

• Brain banks worldwide hold tens of thousands of specimens collected over decades from donors with neurodegenerative diseases, psychiatric conditions, and matched controls
• Whole-cell viability in these samples is zero — enzymatic dissociation is not an option
• Nuclei remain intact and transcriptionally informative for years when tissue is stored at −80 °C
• snRNA-seq opens these archives to systematic molecular investigation

3. Reduced Transcriptional Stress Artifacts

Enzymatic dissociation at 37 °C induces a well-characterized transcriptional stress response dominated by:

• Immediate early genes: FOS, JUN
• Heat-shock proteins: HSPA1A, HSPB1
• Chemokines: CCL3, CCL4

This response is particularly pronounced in microglia and can be mistaken for disease-associated activation. In snRNA-seq, because nuclei are isolated by mechanical means on ice rather than by enzyme incubation, this artifact is largely absent.

Important caveat — postmortem handling artifacts: A systematic dissection of artifactual glial signatures in human postmortem brain by Marsh et al. (2022) [1] found that:

• The enrichment of stress-responsive transcripts in microglia and astrocytes showed no correlation with postmortem interval (PMI) across 49 samples
• Instead, it was linked to donor age, agonal state, and tissue-handling variables that are often unreported in public datasets
• The recommended approach is orthogonal validation — confirm key transcriptional findings with RNA in situ hybridization, spatial transcriptomics, or independent replication — rather than computationally removing these signatures, because the same stress pathways can be genuinely activated in disease

Key takeaway: snRNA-seq from postmortem brain is feasible across a wide PMI range, but careful metadata reporting and orthogonal validation are essential for drawing reliable biological conclusions.

What Frozen and Postmortem Brain Tissue Can Deliver

Frozen and postmortem brain tissue are not simply "acceptable" inputs for snRNA-seq — they are, in practice, the standard starting material for most large-scale human brain atlases published in the past five years. The key variable is not whether the tissue was frozen, but how quickly and uniformly freezing occurred after collection.

Tissue Quality Determinants

A recent benchmarking study comparing snRNA-seq methods on human postmortem brain found that nuclei yield, purity, and library complexity depend more on the choice of isolation protocol than on PMI alone [2].

Nuclei Isolation Workflow

The standard protocol for brain tissue has been refined considerably:

1. Mechanical dissociation of flash-frozen tissue in ice-cold lysis buffer
2. Dounce homogenization with calibrated clearance pestles
3. Filtration through 40 μm strainers to remove debris
4. Optional density gradient centrifugation to enrich for intact nuclei

A simplified version developed specifically for long-term frozen brain tumors (Ernst et al. 2025) [3] reduces the protocol to four steps performed in under 30 minutes:

1. Tissue cutting in lysis buffer
2. Glass Douncing
3. Sequential filtration through 100 μm and 40 μm filters
4. Two to three wash steps

Wash optimization:

• Three washes → cleanest nuclei suspension, lowest debris carryover
• Two washes → preferable for very small tissue inputs where every nucleus counts
• Protocol validated on tissue stored at −80 °C for 6–7 years

Practical Input Requirements

A 20–50 mg piece of frozen brain tissue is typically sufficient to generate a snRNA-seq library on the 10x Genomics platform. The main sample-level requirement is straightforward: the tissue must not have undergone repeated freeze-thaw cycles.

Figure 1: Frozen brain tissue processed for snRNA-seq nuclei isolation — mechanical dissociation in ice-cold buffer and sequential filtration preserve nuclear integrity while removing debris.

Neurons, Glia, and Everything Between

One of the defining findings from brain snRNA-seq studies is the sheer diversity of recoverable cell types — and the fact that non-neuronal populations often outnumber neurons by a wide margin.

Cell-Type Composition: The Substantia Nigra Example

A recent snRNA-seq atlas of the aged human substantia nigra profiled 315,867 nuclei from 32 donors and identified nine distinct cell populations [4]:

Notable discoveries from this dataset:

• A previously undescribed population of RIT2-enriched neurons that is spatially concentrated in the substantia nigra and appears selectively vulnerable in Parkinson's disease
• Approximately 93% of these RIT2-expressing neurons do not express tyrosine hydroxylase (the canonical marker of dopaminergic identity), meaning they would be missed entirely by immunohistochemistry-based approaches
• Within the neuron compartment alone, the study resolved multiple subtypes — dopaminergic neurons (at least two molecularly distinct populations), GABAergic neurons, and glutamatergic neurons

Figure 2: Brain snRNA-seq captures the full spectrum of CNS cell types — neurons, glia, and vascular cells — from a single tissue sample. Non-neuronal populations (oligodendrocytes, microglia, astrocytes) often outnumber neurons, highlighting the importance of unbiased nuclei isolation for comprehensive cell-type profiling.

Neuronal Subtype Recovery

Cortical snRNA-seq routinely captures:

• Excitatory projection neurons: across layers II through VI
• Multiple interneuron subtypes: somatostatin-, parvalbumin-, and VIP-expressing
• Full non-neuronal repertoire: layer-specific astrocytes, homeostatic and activated microglia, oligodendrocytes at different maturation stages, vascular and perivascular cell types

Cross-species comparisons spanning human, chimpanzee, gorilla, macaque, and marmoset have further shown that certain cell-type proportions and gene-expression programs are primate-specific, reinforcing the value of human brain snRNA-seq for translational neuroscience [5].

Microglial Recovery: Resolving the Conflicting Literature

For current platforms, researchers should expect robust microglial representation, provided that:

1. Nuclei isolation conditions remain cold throughout
2. Processing times are kept under one hour from tissue to final suspension

Neurodegeneration Under the Single-Nucleus Lens

Neurodegenerative diseases — where specific neuron populations degenerate while neighboring cell types respond and remodel — are natural targets for snRNA-seq atlasing. Three diseases illustrate the range of what these approaches can reveal.

Alzheimer's Disease

snRNA-seq of postmortem prefrontal cortex from AD donors and age-matched controls has yielded several convergent findings:

Neuron-subtype-specific vulnerabilities:

• Selective vulnerability of somatostatin-expressing interneurons
• Layer-specific excitatory neurons showing transcriptional changes that precede frank neurodegeneration

Disease-associated glial states:

• Microglial population: Enriched for TREM2, APOE, and CD74 — tracks spatially with amyloid pathology
• Astrocyte population: Marked by GFAP and SERPINA3 — localizes to the periphery of amyloid plaques

These disease-associated states are now sufficiently well characterized that they can serve as reference signatures for evaluating new datasets [6]. The ssREAD database further integrates single-cell and spatial transcriptomic data from dozens of AD studies, providing a curated framework for mapping new snRNA-seq profiles against established disease signatures [7].

A comprehensive analysis of multicellular communities in the aging human brain revealed coordinated perturbations across neuronal, glial, and endothelial subpopulations [8]:

• Two multicellular communities were altered in relation to cognitive decline and tau pathology
• The study assembled a high-resolution cellular map of the aging frontal cortex using snRNA-seq of 24 individuals with a range of clinicopathologic characteristics

Parkinson's Disease

The substantia nigra snRNA-seq atlas described above (Wang et al. 2024) [4] identified:

Cell-type-specific findings:

• A novel RIT2-enriched neuron population that is disproportionately lost in PD — validated across two independent published datasets and in human midbrain organoids
• Widespread transcriptomic changes beyond the neuron compartment

Pathway-level disruptions:

These pathway-level disruptions suggest that PD pathology affects not only individual cell types but also the communication networks that coordinate them — a dimension that bulk tissue analysis cannot capture.

Amyotrophic Lateral Sclerosis (ALS)

Two complementary snRNA-seq studies provide a multi-angle view of ALS pathology:

Study 1 — Single-nucleus multiome (Takeuchi et al. 2025) [9]:

• Combined snRNA-seq with snATAC-seq to profile both transcriptomic and chromatin-accessibility changes
• Sampled postmortem motor cortex and spinal cord from ALS donors
• Motor neurons: Pronounced transcriptional dysregulation, including signatures of glutamate overactivation
• Microglia and astrocytes: Adopted disease-associated states that differed between brain and spinal cord — regional specificity with implications for therapeutic targeting

Study 2 — Orbitofrontal cortex atlas (McKeever et al. 2025) [10]:

• 103,076 nuclei from ALS postmortem tissue
• Catalogued transcriptomic shifts across all major cell types
• Identified shared and cell-type-specific gene-expression programs distinguishing ALS from control tissue
• Deep learning-based decoding of alternative polyadenylation mechanisms

Cross-Disease Lesson

The most informative snRNA-seq studies are not those that simply list differentially expressed genes per cluster, but those that contextualize cell-type-specific changes within the broader multicellular network. The technology is well suited to this systems-level view, provided that the study design includes sufficient biological replicates and the analysis moves beyond cluster-level statistics to examine coordinated programs across cell types.

Figure 3: Brain snRNA-seq reveals cell-type-specific transcriptomic signatures across three major neurodegenerative diseases — Alzheimer's disease, Parkinson's disease, and ALS — each affecting distinct neuron populations and glial responses.

Brain Tumors Profiled Nucleus by Nucleus

Brain tumor snRNA-seq sits at the intersection of two challenges: preserving the full spectrum of tumor-intrinsic transcriptional heterogeneity and capturing the tumor microenvironment (TME) in a tissue type where the normal reference — the surrounding brain parenchyma — is itself extraordinarily complex.

Frozen Tumor Protocol: Key Capabilities

The simplified nuclear isolation protocol developed by Ernst et al. (2025) [3] for long-term frozen pediatric brain tumors demonstrated that snRNA-seq can resolve tumor cell populations, infiltrating immune cells, and stromal components from tissue stored for up to seven years at −80 °C.

In pilocytic astrocytoma (low-grade glioma driven by KIAA1549::BRAF fusions):

In pleomorphic xanthoastrocytoma (higher-grade):

• Additional malignant populations emerged: neural-progenitor-like and mesenchymal-like cells
• Mirrors the increased transcriptional heterogeneity associated with glioma progression

Multi-Omic Approaches

Simultaneous snRNA-seq and snATAC-seq from paired tumor core and peripheral regions of glioblastoma (Wang et al. 2024) [11] revealed:

• Chromatin-accessibility landscapes that distinguish invasive-edge tumor cells from proliferative-core populations
• Regulatory elements driving the mesenchymal transcriptional program associated with treatment resistance
• The ability to layer chromatin accessibility onto transcriptomic identity at single-nucleus resolution is particularly valuable in tumors, where epigenetic remodeling often precedes transcriptional changes

Logistical Advantages

snRNA-seq's tolerance for frozen specimens enables:

• Multicenter brain tumor studies where specimens are collected at different surgical centers and shipped frozen to a central processing site
• Re-interrogation of archived tumor collections assembled over years of clinical care using contemporary genomic tools
• These logistical realities would exclude samples from scRNA-seq workflows, which require viable cells

Picking the Right Brain Region

Brain region selection is not a logistical afterthought in snRNA-seq — it is a scientific decision with direct consequences for cell-type recovery, annotation accuracy, and the interpretability of results.

Reference Atlas Availability by Region

Practical Guidelines

1. Match the region to the question: A study of Parkinson's disease pathophysiology requires substantia nigra, not prefrontal cortex — even if cortex would be easier to process and annotate
2. Budget for annotation: Cortical reference atlases are unsuitable for annotating subcortical data. Investigators studying subcortical regions should budget for more extensive marker-gene validation and, where feasible, generate their own reference data from control tissue of the same region
3. Expect yield differences: Cortical gray matter generally gives the highest nuclei counts per milligram. Brainstem and white-matter-enriched regions yield fewer nuclei and may require larger tissue inputs or additional Douncing strokes

Annotating Brain Cell Types With Confidence

Cell-type annotation in brain snRNA-seq is simultaneously easier and harder than in most other tissues.

Why it's easier: The brain is the most extensively catalogued organ in single-cell genomics. Reference datasets span dozens of brain regions and hundreds of transcriptionally defined cell types [6].

Why it's harder: Brain cell types exist along continuous transcriptional gradients — particularly among cortical interneurons and glial cells — making discrete cluster boundaries somewhat artificial. Two cells assigned to different clusters may represent:

• Distinct cell types
• Different states of the same type
• Points along a shared continuum

Standard clustering algorithms cannot distinguish among these possibilities without additional biological context.

Recommended Annotation Tools

Important: For subcortical regions or disease-state samples, reference-based annotation requires more caution. Disease-associated cell states — such as the activated microglial populations seen in AD and ALS, or the reactive astrocyte states identified in multiple neurodegenerative conditions — may not have close counterparts in healthy reference atlases and can be misassigned to the nearest healthy cluster with misleadingly high confidence scores.

Two Practices That Improve Annotation Reliability

1. Validate automated annotations with canonical marker genes:

Marker-based confirmation adds a layer of biological grounding that automated tools — which optimize for statistical similarity rather than biological identity — cannot provide.

2. Treat reference-based labels as hypotheses when studying disease tissue:

A microglial cluster enriched for TREM2 and APOE in an AD sample may map computationally to "homeostatic microglia" in a healthy reference, but its biology is clearly different. The annotation — and any downstream interpretation — should reflect that distinction.

What snRNA-Seq Cannot Tell You

For all its utility, brain snRNA-seq has defined limits that researchers should understand before designing a study. These are not reasons to avoid the technique — they are boundary conditions that, when respected, prevent overinterpretation.

Limitation 1: Nuclear vs. Cytoplasmic Transcriptomes

Implication: Expression values from snRNA-seq and scRNA-seq datasets are not directly interchangeable. Gene-level comparisons between nuclear and whole-cell data require careful normalization. These differences are systematic and predictable — they do not invalidate snRNA-seq.

Limitation 2: Loss of Spatial Context

snRNA-seq dissociates tissue into a single-nucleus suspension, erasing the spatial relationships that define brain circuit function. Two neurons that are synaptically connected in vivo become indistinguishable from two neurons located in opposite hemispheres once nuclei are suspended in buffer.

Solution — paired snRNA-seq + spatial transcriptomics:

• Spatial transcriptomics methods — including those available through CD Genomics — can recover spatial context
• The ssREAD database integrates both modalities for Alzheimer's disease and provides a useful reference framework [7]
• Many groups now design studies with matched snRNA-seq and spatial transcriptomics from adjacent tissue sections
• For neuroscience questions where cell location is essential to biological interpretation, this paired approach is becoming the field standard

Limitation 3: Temporal Snapshot vs. Disease Progression

The postmortem or perioperative brain tissue used for snRNA-seq captures a single transcriptional snapshot — usually the end state of a disease process that unfolded over decades.

Implication: Transcriptional changes observed in postmortem AD or PD brain reflect the culmination of pathology, not its initiation. Mechanistic inference about disease progression requires complementary approaches, including model systems where the temporal trajectory can be studied longitudinally and compared against the postmortem endpoint.

Limitation 4: Artifactual Glial Signatures

As noted in the first section, artifactual glial activation signatures are a genuine interpretive hazard that does not disappear with larger sample sizes or deeper sequencing. Marsh et al. [1] explicitly caution against computationally removing these signatures, because:

• The same stress pathways can be genuinely activated in disease
• Regressing them out risks discarding real biology along with artifact
• Recommended approach: orthogonal validation — confirm key transcriptional findings with RNA in situ hybridization, spatial transcriptomics, or independent replication in cohorts with documented tissue-handling metadata

Limitations at a Glance

Researchers planning brain snRNA-seq studies can explore our snRNA sequencing services for experimental design support, from nuclei isolation to data analysis. For projects that pair snRNA-seq with spatial context, our spatial transcriptomics services offer complementary tissue-level readouts.

FAQ

Q: Can snRNA-seq from postmortem brain tissue recover all major cell types equally?

A: snRNA-seq from frozen or postmortem brain consistently recovers all major cell types in the central nervous system, including the full neuronal repertoire — excitatory projection neurons across all cortical layers and multiple interneuron subtypes — with substantially better neuronal representation than scRNA-seq, where enzymatic dissociation selectively destroys large, morphologically complex neurons. Non-neuronal populations are also consistently recovered: astrocytes, oligodendrocytes (which dominate most brain regions numerically), microglia (robustly detected with 10x Genomics v3 chemistry and later, including activation-related genes such as SPP1, CD74, and APOE), oligodendrocyte progenitor cells, endothelial cells, and pericytes. For optimal recovery across all cell types, nuclei isolation should be kept cold throughout the entire protocol, total processing time from tissue to final suspension should be kept under one hour, and current 10x Genomics chemistry (v3 or later) should be used.

Q: How long after death can brain tissue be used for snRNA-seq?

A: Postmortem interval (PMI) is not the dominant driver of data quality in brain snRNA-seq. A systematic examination of 49 human postmortem brain samples by Marsh et al. (2022) [1] found no correlation between stress-responsive transcript enrichment and PMI. Tissue that was rapidly collected and flash-frozen within hours of death produced analyzable data even when PMI extended beyond 24 hours. The variables that matter more are donor age, agonal state, and tissue-handling conditions before freezing rather than the absolute time since death. For surgical resections, snap-freezing within 30 minutes of excision is ideal. For banked postmortem specimens, consistent −80 °C storage without freeze-thaw cycles is more important than the absolute PMI value.

Q: How many nuclei should I target for a brain snRNA-seq experiment?

A: The recommended nuclei count depends on the study objective. For standard cell-type profiling — sufficient to resolve major cell types and their transcriptional states — 5,000–10,000 nuclei per sample is adequate. For detecting rare cell populations such as disease-associated microglial subtypes or specific interneuron classes, 15,000–20,000 nuclei per sample is recommended. For building a single-region, single-condition reference atlas, approximately 50,000 nuclei is a reasonable target. For multi-donor, multi-region atlases, substantially larger datasets are required — the substantia nigra study by Wang et al. (2024) [4] profiled 315,867 nuclei from 32 donors. Regarding input tissue, a 20–50 mg piece of frozen brain tissue is usually sufficient to generate 10,000 nuclei, though yields vary by region: cortical gray matter produces more nuclei per milligram than brainstem or white-matter tracts.

Q: Can I use the same cell-type annotation reference for different brain regions?

A: No — and this is one of the most common pitfalls in brain snRNA-seq analysis. Different brain regions have fundamentally different cellular compositions, so a reference atlas built for one region will misannotate data from another. For cortical studies, the Allen Brain Atlas, Human Cell Atlas, and BRAIN Initiative Cell Census Network provide well-annotated references spanning hundreds of cell types. For hippocampus, the integrated snRNA-seq and spatial transcriptomics atlas by Thompson et al. (2025) [12] is the appropriate reference — not cortical atlases. For substantia nigra, region-matched reference data is essential because the cellular composition differs dramatically from cortex (approximately 51% oligodendrocytes versus ~13% neurons, compared with far higher neuronal proportions in cortical samples). For thalamus, hypothalamus, striatum, and brainstem — where published reference datasets are limited — investigators should budget for region-matched reference data, validate automated annotations with canonical marker genes, and consider generating their own reference data from control tissue of the same region processed in parallel.

Q: Should I pair snRNA-seq with spatial transcriptomics for my brain study?

A: It depends entirely on the research question. For studies focused on cell-type discovery, transcriptional state characterization, or gene-regulatory network analysis, snRNA-seq alone may be sufficient. However, when the biological question involves spatial relationships — such as mapping disease-associated glial states relative to amyloid plaques in Alzheimer's disease, or identifying which neuronal populations communicate across cortical layers — pairing snRNA-seq with spatial transcriptomics becomes essential. For studies where tissue-level architecture must be combined with deep transcriptomic profiling, the paired approach is increasingly the field standard. The ssREAD database [7] integrates both modalities for Alzheimer's disease and provides a useful reference framework for combined analyses.

References

1. Marsh SE, Walker AJ, Kamath T, et al. Dissection of artifactual and confounding glial signatures by single-cell sequencing of mouse and human brain. Nature Neuroscience. 2022;25:306-316. doi:10.1038/s41593-022-01022-8
2. Benchmarking of single nuclei RNA-seq methods on human post-mortem brain tissue. Genomics. 2026;118(1):111184. doi:10.1016/j.ygeno.2025.111184
3. Ernst KJ, Okonechnikov K, Bageritz J, et al. A simplified preparation method for single-nucleus RNA-sequencing using long-term frozen brain tumor tissues. Scientific Reports. 2025;15:12849. doi:10.1038/s41598-025-97053-9
4. Wang Q, Wang M, Choi I, et al. Molecular profiling of human substantia nigra identifies diverse neuron types associated with Parkinson's disease. Science Advances. 2024;10(2):eadi8287. doi:10.1126/sciadv.adi8287
5. Jorstad NL, Close J, Johansen N, et al. Comparative transcriptomics reveals human-specific cortical features. Science. 2023;382(6667):eade9516. doi:10.1126/science.ade9516
6. Bonev B, Castelo-Branco G, Chen S, et al. Opportunities and challenges of single-cell and spatially resolved genomics methods for neuroscience discovery. Nature Neuroscience. 2024;27(12):2292-2309. doi:10.1038/s41593-024-01806-0
7. Wang C, Acosta D, McNutt M, et al. A single-cell and spatial RNA-seq database for Alzheimer's disease (ssREAD). Nature Communications. 2024;15:4710. doi:10.1038/s41467-024-49133-z
8. Cain A, Taga M, McCabe C, et al. Multicellular communities are perturbed in the aging human brain and Alzheimer's disease. Nature Neuroscience. 2023;26:1267-1280. doi:10.1038/s41593-023-01356-x
9. Takeuchi E, Yasumizu Y, Morita J, et al. Single-nucleus multiome shows motor neuron glutamate overactivation in amyotrophic lateral sclerosis. Brain. 2025. doi:10.1093/brain/awaf426
10. McKeever PM, Sababi AM, Sharma R, et al. Single-nucleus transcriptome atlas of orbitofrontal cortex in ALS with a deep learning-based decoding of alternative polyadenylation mechanisms. Cell Genomics. 2025;5(12):101007. doi:10.1016/j.xgen.2025.101007
11. Wang X, Sun Q, Liu T, et al. Single-cell multi-omics sequencing uncovers region-specific plasticity of glioblastoma for complementary therapeutic targeting. Science Advances. 2024;10(47):eadn4306. doi:10.1126/sciadv.adn4306
12. Thompson JR, Nelson ED, Tippani M, et al. An integrated single-nucleus and spatial transcriptomics atlas reveals the molecular landscape of the human hippocampus. Nature Neuroscience. 2025;28:1990-2004. doi:10.1038/s41593-025-02022-0